www.cmm.bristol.ac.uk/forum

Hi all
I am running a 2-level model with about 60 higher level units which have been randomly sampled from a population of about 150; thus about 40% of the entire population of higher level units is included in the sample. Each higher level unit (within and outwith the sample) is approximately the same size, with about 1000 lower level units clustered within each.

All lower level units within each higher level unit in the sample are included in the sample; i.e. the sample is a cluster sample. There are covariates at both levels.

I'm familiar with the concept of applying a correction for finite populations to standard errors in single level regression models, but not sure how to proceed with a multilevel model. I wonder should the correction be applied only to higher level variables, or to variables at both levels.

Apologies if this seems not a very sensible question: I don't have much experience with MLwiN or multilevel modelling.

Any advice gratefully appreciated.
John

Hi John,
There isn't a whole lot of literature on FPCs for multilevel models and MLwiN assumes an infinite population in all it's calculations. A search will find a few recent papers if you are interested.
Best wishes,
Bill.

Thanks Bill.
The papers I have read are a little inconsistent and/or vague on this matter, but on the whole seem to consider that correcting at both levels is appropriate. I'd be interested to hear from anyone who has faced the same issue or arrived at similar or different conclusions.
Best wishes
John

Hi John, what papers have you read on this? It might help us to understand the matter a bit more.